FAILED: patch "[PATCH] ext4: fix potential race between online resizing and write" failed to apply to 4.14-stable tree

PATCH In review [email protected] 3 participants 5 messages Feb 27, 2020 – Feb 29, 2020

Subsystems (2)

Stable branch ext4 file system

lore

Share & tools

lore

[email protected]

author maintainer

2020-02-27 08:08 UTC #0 lore raw

The patch below does not apply to the 4.14-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 1d0c3924a92e69bfa91163bda83c12a994b4d106 Mon Sep 17 00:00:00 2001
From: Theodore Ts'o <[email protected]>
Date: Sat, 15 Feb 2020 16:40:37 -0500
Subject: [PATCH] ext4: fix potential race between online resizing and write
operations

During an online resize an array of pointers to buffer heads gets replaced so it can get enlarged. If there is a racing block allocation or deallocation which uses the old array, and the old array has gotten reused this can lead to a GPF or some other random kernel memory getting modified.

Link https://bugzilla.kernel.org/show_bug.cgi?id=206443

Link https://lore.kernel.org/r/[email protected] open here

Reported-by Suraj Jitindar Singh <[email protected]>

Signed-off-by Theodore Ts'o <[email protected]>

Cc [email protected]

fs/ext4/balloc.c +11 -3

index 5f993a411251..8fd0b3cdab4c 100644
@@ -270,6 +270,7 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
 	ext4_group_t ngroups = ext4_get_groups_count(sb);
 	struct ext4_group_desc *desc;
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
+	struct buffer_head *bh_p;
 
 	if (block_group >= ngroups) {
 		ext4_error(sb, "block_group >= groups_count - block_group = %u,"

@@ -280,7 +281,14 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
 
 	group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
 	offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
-	if (!sbi->s_group_desc[group_desc]) {
+	bh_p = sbi_array_rcu_deref(sbi, s_group_desc, group_desc);
+	/*
+	 * sbi_array_rcu_deref returns with rcu unlocked, this is ok since
+	 * the pointer being dereferenced won't be dereferenced again. By
+	 * looking at the usage in add_new_gdb() the value isn't modified,
+	 * just the pointer, and so it remains valid.
+	 */
+	if (!bh_p) {
 		ext4_error(sb, "Group descriptor not loaded - "
 			   "block_group = %u, group_desc = %u, desc = %u",
 			   block_group, group_desc, offset);

@@ -288,10 +296,10 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
 	}
 
 	desc = (struct ext4_group_desc *)(
-		(__u8 *)sbi->s_group_desc[group_desc]->b_data +
+		(__u8 *)bh_p->b_data +
 		offset * EXT4_DESC_SIZE(sb));
 	if (bh)
-		*bh = sbi->s_group_desc[group_desc];
+		*bh = bh_p;
 	return desc;
 }

fs/ext4/ext4.h +19 -1

index 480badcf2783..b51003f75568 100644
@@ -1400,7 +1400,7 @@ struct ext4_sb_info {
 	loff_t s_bitmap_maxbytes;	/* max bytes for bitmap files */
 	struct buffer_head * s_sbh;	/* Buffer containing the super block */
 	struct ext4_super_block *s_es;	/* Pointer to the super block in the buffer */
-	struct buffer_head **s_group_desc;
+	struct buffer_head * __rcu *s_group_desc;
 	unsigned int s_mount_opt;
 	unsigned int s_mount_opt2;
 	unsigned int s_mount_flags;

@@ -1576,6 +1576,23 @@ static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
 		 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
 }
 
+/*
+ * Returns: sbi->field[index]
+ * Used to access an array element from the following sbi fields which require
+ * rcu protection to avoid dereferencing an invalid pointer due to reassignment
+ * - s_group_desc
+ * - s_group_info
+ * - s_flex_group
+ */
+#define sbi_array_rcu_deref(sbi, field, index)				   \
+({									   \
+	typeof(*((sbi)->field)) _v;					   \
+	rcu_read_lock();						   \
+	_v = ((typeof(_v)*)rcu_dereference((sbi)->field))[index];	   \
+	rcu_read_unlock();						   \
+	_v;								   \
+})
+
 /*
  * Simulate_fail codes
  */

@@ -2730,6 +2747,7 @@ extern int ext4_generic_delete_entry(handle_t *handle,
 extern bool ext4_empty_dir(struct inode *inode);
 
 /* resize.c */
+extern void ext4_kvfree_array_rcu(void *to_free);
 extern int ext4_group_add(struct super_block *sb,
 				struct ext4_new_group_data *input);
 extern int ext4_group_extend(struct super_block *sb,

fs/ext4/resize.c +44 ~~-11~~

110 lines not on this page · raw message

fs/ext4/super.c +23 ~~-10~~

index f464dff09774..e00bcc19099f 100644
@@ -1014,6 +1014,7 @@ static void ext4_put_super(struct super_block *sb)
 {
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
 	struct ext4_super_block *es = sbi->s_es;
+	struct buffer_head **group_desc;
 	int aborted = 0;
 	int i, err;

@@ -1046,9 +1047,12 @@ static void ext4_put_super(struct super_block *sb)
 	if (!sb_rdonly(sb))
 		ext4_commit_super(sb, 1);
 
+	rcu_read_lock();
+	group_desc = rcu_dereference(sbi->s_group_desc);
 	for (i = 0; i < sbi->s_gdb_count; i++)
-		brelse(sbi->s_group_desc[i]);
-	kvfree(sbi->s_group_desc);
+		brelse(group_desc[i]);
+	kvfree(group_desc);
+	rcu_read_unlock();
 	kvfree(sbi->s_flex_groups);
 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
 	percpu_counter_destroy(&sbi->s_freeinodes_counter);

@@ -3634,7 +3638,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 {
 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 	char *orig_data = kstrdup(data, GFP_KERNEL);
-	struct buffer_head *bh;
+	struct buffer_head *bh, **group_desc;
 	struct ext4_super_block *es = NULL;
 	struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 	ext4_fsblk_t block;

@@ -4290,9 +4294,10 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 			goto failed_mount;
 		}
 	}
-	sbi->s_group_desc = kvmalloc_array(db_count,
-					   sizeof(struct buffer_head *),
-					   GFP_KERNEL);
+	rcu_assign_pointer(sbi->s_group_desc,
+			   kvmalloc_array(db_count,
+					  sizeof(struct buffer_head *),
+					  GFP_KERNEL));
 	if (sbi->s_group_desc == NULL) {
 		ext4_msg(sb, KERN_ERR, "not enough memory");
 		ret = -ENOMEM;

@@ -4308,14 +4313,19 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	}
 
 	for (i = 0; i < db_count; i++) {
+		struct buffer_head *bh;
+
 		block = descriptor_loc(sb, logical_sb_block, i);
-		sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
-		if (!sbi->s_group_desc[i]) {
+		bh = sb_bread_unmovable(sb, block);
+		if (!bh) {
 			ext4_msg(sb, KERN_ERR,
 			       "can't read group descriptor %d", i);
 			db_count = i;
 			goto failed_mount2;
 		}
+		rcu_read_lock();
+		rcu_dereference(sbi->s_group_desc)[i] = bh;
+		rcu_read_unlock();
 	}
 	sbi->s_gdb_count = db_count;
 	if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {

@@ -4717,9 +4727,12 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	if (sbi->s_mmp_tsk)
 		kthread_stop(sbi->s_mmp_tsk);
 failed_mount2:
+	rcu_read_lock();
+	group_desc = rcu_dereference(sbi->s_group_desc);
 	for (i = 0; i < db_count; i++)
-		brelse(sbi->s_group_desc[i]);
-	kvfree(sbi->s_group_desc);
+		brelse(group_desc[i]);
+	kvfree(group_desc);
+	rcu_read_unlock();
 failed_mount:
 	if (sbi->s_chksum_driver)
 		crypto_free_shash(sbi->s_chksum_driver);

Suraj Jitindar Singh

2020-02-29 00:48 UTC #1 lore raw

PATCH 1/3 ext4: fix potential race between online resizing and write operations

From: Theodore Ts'o <[email protected]>

commit 1d0c3924a92e69bfa91163bda83c12a994b4d106 upstream.

Link https://bugzilla.kernel.org/show_bug.cgi?id=206443

Link https://lore.kernel.org/r/[email protected] open here

Reported-by Suraj Jitindar Singh <[email protected]>

Signed-off-by Theodore Ts'o <[email protected]>

Cc [email protected] # 4.14.x

---

Changes against upstream:
fs/ext4/ext4.h:
- missing context, massage to apply
- no functional change
fs/ext4/super.c:ext4_fill_super()
- kvmalloc converted to kvmalloc_array upstream
- convert to use kvmalloc_array

---
 fs/ext4/balloc.c | 14 +++++++++---
 fs/ext4/ext4.h   | 20 +++++++++++++++++-
 fs/ext4/resize.c | 55 ++++++++++++++++++++++++++++++++++++++----------
 fs/ext4/super.c  | 31 +++++++++++++++++++--------
 4 files changed, 96 insertions(+), 24 deletions(-)

fs/ext4/balloc.c +11 -3

index 70266a3355dc..fb38f20f869e 100644
@@ -280,6 +280,7 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
 	ext4_group_t ngroups = ext4_get_groups_count(sb);
 	struct ext4_group_desc *desc;
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
+	struct buffer_head *bh_p;
 
 	if (block_group >= ngroups) {
 		ext4_error(sb, "block_group >= groups_count - block_group = %u,"

@@ -290,7 +291,14 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
 
 	group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
 	offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
-	if (!sbi->s_group_desc[group_desc]) {
+	bh_p = sbi_array_rcu_deref(sbi, s_group_desc, group_desc);
+	/*
+	 * sbi_array_rcu_deref returns with rcu unlocked, this is ok since
+	 * the pointer being dereferenced won't be dereferenced again. By
+	 * looking at the usage in add_new_gdb() the value isn't modified,
+	 * just the pointer, and so it remains valid.
+	 */
+	if (!bh_p) {
 		ext4_error(sb, "Group descriptor not loaded - "
 			   "block_group = %u, group_desc = %u, desc = %u",
 			   block_group, group_desc, offset);

@@ -298,10 +306,10 @@ struct ext4_group_desc * ext4_get_group_desc(struct super_block *sb,
 	}
 
 	desc = (struct ext4_group_desc *)(
-		(__u8 *)sbi->s_group_desc[group_desc]->b_data +
+		(__u8 *)bh_p->b_data +
 		offset * EXT4_DESC_SIZE(sb));
 	if (bh)
-		*bh = sbi->s_group_desc[group_desc];
+		*bh = bh_p;
 	return desc;
 }

fs/ext4/ext4.h +19 -1

index 95ef26b39e69..0c3b550f9822 100644
@@ -1382,7 +1382,7 @@ struct ext4_sb_info {
 	loff_t s_bitmap_maxbytes;	/* max bytes for bitmap files */
 	struct buffer_head * s_sbh;	/* Buffer containing the super block */
 	struct ext4_super_block *s_es;	/* Pointer to the super block in the buffer */
-	struct buffer_head **s_group_desc;
+	struct buffer_head * __rcu *s_group_desc;
 	unsigned int s_mount_opt;
 	unsigned int s_mount_opt2;
 	unsigned int s_mount_flags;

@@ -1553,6 +1553,23 @@ static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
 		 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
 }
 
+/*
+ * Returns: sbi->field[index]
+ * Used to access an array element from the following sbi fields which require
+ * rcu protection to avoid dereferencing an invalid pointer due to reassignment
+ * - s_group_desc
+ * - s_group_info
+ * - s_flex_group
+ */
+#define sbi_array_rcu_deref(sbi, field, index)				   \
+({									   \
+	typeof(*((sbi)->field)) _v;					   \
+	rcu_read_lock();						   \
+	_v = ((typeof(_v)*)rcu_dereference((sbi)->field))[index];	   \
+	rcu_read_unlock();						   \
+	_v;								   \
+})
+
 /*
  * Inode dynamic state flags
  */

@@ -2563,6 +2580,7 @@ extern int ext4_generic_delete_entry(handle_t *handle,
 extern bool ext4_empty_dir(struct inode *inode);
 
 /* resize.c */
+extern void ext4_kvfree_array_rcu(void *to_free);
 extern int ext4_group_add(struct super_block *sb,
 				struct ext4_new_group_data *input);
 extern int ext4_group_extend(struct super_block *sb,

fs/ext4/resize.c +44 ~~-11~~

110 lines not on this page · raw message

fs/ext4/super.c +22 -9

index 93d8aa6ef661..3d4e8a2a1f7a 100644
@@ -900,6 +900,7 @@ static void ext4_put_super(struct super_block *sb)
 {
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
 	struct ext4_super_block *es = sbi->s_es;
+	struct buffer_head **group_desc;
 	int aborted = 0;
 	int i, err;

@@ -931,9 +932,12 @@ static void ext4_put_super(struct super_block *sb)
 	if (!sb_rdonly(sb))
 		ext4_commit_super(sb, 1);
 
+	rcu_read_lock();
+	group_desc = rcu_dereference(sbi->s_group_desc);
 	for (i = 0; i < sbi->s_gdb_count; i++)
-		brelse(sbi->s_group_desc[i]);
-	kvfree(sbi->s_group_desc);
+		brelse(group_desc[i]);
+	kvfree(group_desc);
+	rcu_read_unlock();
 	kvfree(sbi->s_flex_groups);
 	percpu_counter_destroy(&sbi->s_freeclusters_counter);
 	percpu_counter_destroy(&sbi->s_freeinodes_counter);

@@ -3495,7 +3499,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 {
 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
 	char *orig_data = kstrdup(data, GFP_KERNEL);
-	struct buffer_head *bh;
+	struct buffer_head *bh, **group_desc;
 	struct ext4_super_block *es = NULL;
 	struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 	ext4_fsblk_t block;

@@ -4108,9 +4112,10 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 			goto failed_mount;
 		}
 	}
-	sbi->s_group_desc = kvmalloc(db_count *
+	rcu_assign_pointer(sbi->s_group_desc,
+			   kvmalloc_array(db_count,
 					  sizeof(struct buffer_head *),
-					  GFP_KERNEL);
+					  GFP_KERNEL));
 	if (sbi->s_group_desc == NULL) {
 		ext4_msg(sb, KERN_ERR, "not enough memory");
 		ret = -ENOMEM;

@@ -4126,14 +4131,19 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	}
 
 	for (i = 0; i < db_count; i++) {
+		struct buffer_head *bh;
+
 		block = descriptor_loc(sb, logical_sb_block, i);
-		sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
-		if (!sbi->s_group_desc[i]) {
+		bh = sb_bread_unmovable(sb, block);
+		if (!bh) {
 			ext4_msg(sb, KERN_ERR,
 			       "can't read group descriptor %d", i);
 			db_count = i;
 			goto failed_mount2;
 		}
+		rcu_read_lock();
+		rcu_dereference(sbi->s_group_desc)[i] = bh;
+		rcu_read_unlock();
 	}
 	sbi->s_gdb_count = db_count;
 	if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {

@@ -4525,9 +4535,12 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	if (sbi->s_mmp_tsk)
 		kthread_stop(sbi->s_mmp_tsk);
 failed_mount2:
+	rcu_read_lock();
+	group_desc = rcu_dereference(sbi->s_group_desc);
 	for (i = 0; i < db_count; i++)
-		brelse(sbi->s_group_desc[i]);
-	kvfree(sbi->s_group_desc);
+		brelse(group_desc[i]);
+	kvfree(group_desc);
+	rcu_read_unlock();
 failed_mount:
 	if (sbi->s_chksum_driver)
 		crypto_free_shash(sbi->s_chksum_driver);

Signature · 1 line

2.17.1

Suraj Jitindar Singh

2020-02-29 00:48 UTC #2 lore raw

PATCH 2/3 ext4: fix potential race between s_group_info online resizing and access

commit df3da4ea5a0fc5d115c90d5aa6caa4dd433750a7 upstream.

During an online resize an array of pointers to s_group_info gets replaced so it can get enlarged. If there is a concurrent access to the array in ext4_get_group_info() and this memory has been reused then this can lead to an invalid memory access.

Link https://bugzilla.kernel.org/show_bug.cgi?id=206443

Link https://lore.kernel.org/r/[email protected] open here

Signed-off-by Suraj Jitindar Singh <[email protected]>

Signed-off-by Theodore Ts'o <[email protected]>

Reviewed-by Balbir Singh <[email protected]>

Cc [email protected] # 4.14.x

---

Changes against upstream:
none

---
 fs/ext4/ext4.h    |  8 ++++----
 fs/ext4/mballoc.c | 52 +++++++++++++++++++++++++++++++----------------
 2 files changed, 39 insertions(+), 21 deletions(-)

fs/ext4/ext4.h +4 -4

index 0c3b550f9822..ef3204caf2da 100644
@@ -1442,7 +1442,7 @@ struct ext4_sb_info {
 #endif
 
 	/* for buddy allocator */
-	struct ext4_group_info ***s_group_info;
+	struct ext4_group_info ** __rcu *s_group_info;
 	struct inode *s_buddy_cache;
 	spinlock_t s_md_lock;
 	unsigned short *s_mb_offsets;

@@ -2826,13 +2826,13 @@ static inline
 struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
 					    ext4_group_t group)
 {
-	 struct ext4_group_info ***grp_info;
+	 struct ext4_group_info **grp_info;
 	 long indexv, indexh;
 	 BUG_ON(group >= EXT4_SB(sb)->s_groups_count);
-	 grp_info = EXT4_SB(sb)->s_group_info;
 	 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
 	 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
-	 return grp_info[indexv][indexh];
+	 grp_info = sbi_array_rcu_deref(EXT4_SB(sb), s_group_info, indexv);
+	 return grp_info[indexh];
 }
 
 /*

fs/ext4/mballoc.c +35 ~~-17~~

114 lines not on this page · raw message

Signature · 1 line

2.17.1

Suraj Jitindar Singh

2020-02-29 00:48 UTC #3 lore raw

PATCH 3/3 ext4: fix potential race between s_flex_groups online resizing and access

In reply to Suraj Jitindar Singh

commit 7c990728b99ed6fbe9c75fc202fce1172d9916da upstream.

During an online resize an array of s_flex_groups structures gets replaced so it can get enlarged. If there is a concurrent access to the array and this memory has been reused then this can lead to an invalid memory access.

The s_flex_group array has been converted into an array of pointers rather than an array of structures. This is to ensure that the information contained in the structures cannot get out of sync during a resize due to an accessor updating the value in the old structure after it has been copied but before the array pointer is updated. Since the structures them- selves are no longer copied but only the pointers to them this case is mitigated.

Link https://bugzilla.kernel.org/show_bug.cgi?id=206443

Link https://lore.kernel.org/r/[email protected] open here

Signed-off-by Suraj Jitindar Singh <[email protected]>

Signed-off-by Theodore Ts'o <[email protected]>

Cc [email protected] # 4.14.x

---

Changes against upstream:
fs/ext4/mballoc.c:ext4_group_add_blocks()
- blocks_freed converted to clusters_freed upstream
- preserve blocks_freed usage

---
 fs/ext4/ext4.h    |  2 +-
 fs/ext4/ialloc.c  | 23 +++++++++------
 fs/ext4/mballoc.c |  9 ++++--
 fs/ext4/resize.c  |  7 +++--
 fs/ext4/super.c   | 72 ++++++++++++++++++++++++++++++++---------------
 5 files changed, 76 insertions(+), 37 deletions(-)

fs/ext4/ext4.h +1 -1

index ef3204caf2da..e06a0b818256 100644
@@ -1492,7 +1492,7 @@ struct ext4_sb_info {
 	unsigned int s_extent_max_zeroout_kb;
 
 	unsigned int s_log_groups_per_flex;
-	struct flex_groups *s_flex_groups;
+	struct flex_groups * __rcu *s_flex_groups;
 	ext4_group_t s_flex_groups_allocated;
 
 	/* workqueue for reserved extent conversions (buffered io) */

fs/ext4/ialloc.c +14 -9

index 2f46564d3fca..2a480c0ef1bc 100644
@@ -333,11 +333,13 @@ void ext4_free_inode(handle_t *handle, struct inode *inode)
 
 	percpu_counter_inc(&sbi->s_freeinodes_counter);
 	if (sbi->s_log_groups_per_flex) {
-		ext4_group_t f = ext4_flex_group(sbi, block_group);
+		struct flex_groups *fg;
 
-		atomic_inc(&sbi->s_flex_groups[f].free_inodes);
+		fg = sbi_array_rcu_deref(sbi, s_flex_groups,
+					 ext4_flex_group(sbi, block_group));
+		atomic_inc(&fg->free_inodes);
 		if (is_directory)
-			atomic_dec(&sbi->s_flex_groups[f].used_dirs);
+			atomic_dec(&fg->used_dirs);
 	}
 	BUFFER_TRACE(bh2, "call ext4_handle_dirty_metadata");
 	fatal = ext4_handle_dirty_metadata(handle, NULL, bh2);

@@ -378,12 +380,13 @@ static void get_orlov_stats(struct super_block *sb, ext4_group_t g,
 			    int flex_size, struct orlov_stats *stats)
 {
 	struct ext4_group_desc *desc;
-	struct flex_groups *flex_group = EXT4_SB(sb)->s_flex_groups;
 
 	if (flex_size > 1) {
-		stats->free_inodes = atomic_read(&flex_group[g].free_inodes);
-		stats->free_clusters = atomic64_read(&flex_group[g].free_clusters);
-		stats->used_dirs = atomic_read(&flex_group[g].used_dirs);
+		struct flex_groups *fg = sbi_array_rcu_deref(EXT4_SB(sb),
+							     s_flex_groups, g);
+		stats->free_inodes = atomic_read(&fg->free_inodes);
+		stats->free_clusters = atomic64_read(&fg->free_clusters);
+		stats->used_dirs = atomic_read(&fg->used_dirs);
 		return;
 	}

@@ -1062,7 +1065,8 @@ struct inode *__ext4_new_inode(handle_t *handle, struct inode *dir,
 		if (sbi->s_log_groups_per_flex) {
 			ext4_group_t f = ext4_flex_group(sbi, group);
 
-			atomic_inc(&sbi->s_flex_groups[f].used_dirs);
+			atomic_inc(&sbi_array_rcu_deref(sbi, s_flex_groups,
+							f)->used_dirs);
 		}
 	}
 	if (ext4_has_group_desc_csum(sb)) {

@@ -1085,7 +1089,8 @@ struct inode *__ext4_new_inode(handle_t *handle, struct inode *dir,
 
 	if (sbi->s_log_groups_per_flex) {
 		flex_group = ext4_flex_group(sbi, group);
-		atomic_dec(&sbi->s_flex_groups[flex_group].free_inodes);
+		atomic_dec(&sbi_array_rcu_deref(sbi, s_flex_groups,
+						flex_group)->free_inodes);
 	}
 
 	inode->i_ino = ino + group * EXT4_INODES_PER_GROUP(sb);

fs/ext4/mballoc.c +6 -3

index 4ca540c3b14a..745a89d30a57 100644
@@ -3070,7 +3070,8 @@ ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
 		ext4_group_t flex_group = ext4_flex_group(sbi,
 							  ac->ac_b_ex.fe_group);
 		atomic64_sub(ac->ac_b_ex.fe_len,
-			     &sbi->s_flex_groups[flex_group].free_clusters);
+			     &sbi_array_rcu_deref(sbi, s_flex_groups,
+						  flex_group)->free_clusters);
 	}
 
 	err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);

@@ -4965,7 +4966,8 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
 	if (sbi->s_log_groups_per_flex) {
 		ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
 		atomic64_add(count_clusters,
-			     &sbi->s_flex_groups[flex_group].free_clusters);
+			     &sbi_array_rcu_deref(sbi, s_flex_groups,
+						  flex_group)->free_clusters);
 	}
 
 	if (!(flags & EXT4_FREE_BLOCKS_NO_QUOT_UPDATE))

@@ -5110,7 +5112,8 @@ int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
 	if (sbi->s_log_groups_per_flex) {
 		ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
 		atomic64_add(EXT4_NUM_B2C(sbi, blocks_freed),
-			     &sbi->s_flex_groups[flex_group].free_clusters);
+			     &sbi_array_rcu_deref(sbi, s_flex_groups,
+						  flex_group)->free_clusters);
 	}
 
 	ext4_mb_unload_buddy(&e4b);

fs/ext4/resize.c +5 -2

index 16e3830da548..d42f7471fd5b 100644
@@ -1425,11 +1425,14 @@ static void ext4_update_super(struct super_block *sb,
 		   percpu_counter_read(&sbi->s_freeclusters_counter));
 	if (ext4_has_feature_flex_bg(sb) && sbi->s_log_groups_per_flex) {
 		ext4_group_t flex_group;
+		struct flex_groups *fg;
+
 		flex_group = ext4_flex_group(sbi, group_data[0].group);
+		fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
 		atomic64_add(EXT4_NUM_B2C(sbi, free_blocks),
-			     &sbi->s_flex_groups[flex_group].free_clusters);
+			     &fg->free_clusters);
 		atomic_add(EXT4_INODES_PER_GROUP(sb) * flex_gd->count,
-			   &sbi->s_flex_groups[flex_group].free_inodes);
+			   &fg->free_inodes);
 	}
 
 	/*

fs/ext4/super.c +50 ~~-22~~

126 lines not on this page · raw message

Signature · 1 line

2.17.1

Sasha Levin

maintainer

2020-02-29 03:59 UTC #4 lore raw

Re: [PATCH 4.14.x 1/3] ext4: fix potential race between online resizing and write operations

In reply to Suraj Jitindar Singh

Quoted 10 lines · Suraj Jitindar Singh

From: Theodore Ts'o <[email protected]>
commit 1d0c3924a92e69bfa91163bda83c12a994b4d106 upstream.
During an online resize an array of pointers to buffer heads gets replaced so it can get enlarged. If there is a racing block allocation or deallocation which uses the old array, and the old array has gotten reused this can lead to a GPF or some other random kernel memory getting modified.

Quoted 5 lines

Link https://bugzilla.kernel.org/show_bug.cgi?id=206443
Link https://lore.kernel.org/r/[email protected] open here
Reported-by Suraj Jitindar Singh <[email protected]>
Signed-off-by Theodore Ts'o <[email protected]>
Cc [email protected] # 4.14.x

I've applied this and the 4.9 and 4.4 series.

Note that patch 2 in all of your serieses didn't apply cleanly for me, but cherry picking the upstream commit directly worked so I did that.

Signature · 2 lines

Thanks,
Sasha